As 2023 comes to an end, it marks one more year in the history books. But for quants, this has been no ordinary year.
For the first time in decades, interest rates have rocketed from the ZIRP baseline, a new technology has gripped the landscape, and a new form of options has shown its full-year in action. Put simply, there is no better time to be a quant.
With the year wrapping up, we’ll be touching on who and what strategies are making the most profits, why things are different this time, and where there’s still tons of alpha to be found.
Without further ado, let’s dive right in.
The Players: Who Are They?
For good reasons, many quant firms hold their profit figures tightly, so we have to get a bit creative to figure out what’s going on behind the scenes. To start, let’s pull some alternative data to see who’s hiring and what strategies they’re hiring for:
Taking just Optiver for instance, it seems that they’re rapidly expanding their footprint to quantitative strategies in the commodities space. But before we dive deeper into that, we need to peek inside Optiver’s core business that rakes in the major dough needed for expansion: option market making.
As we briefly discussed, the introduction of 0-DTE contracts in late-’22 has not been just a mere footnote in history. Just a few days prior to this post, CBOE announced that yet another month had seen record volume, boasting a 34.5% increase in index options from the year prior:
With record-breaking amounts of option cash changing hands every day, it’s only natural that middle-manning is more profitable than ever. There are dozens of approaches under the option market making umbrella, but one in which Optiver excels at is mass quoting — let’s look into that.
Mass quoting is just what it sounds like; the act of sending a mass of quotes across venues and underlyings simultaneously. To first be able to provide an active quote, the market maker generally prices the option with a traditional model (e.g., Black-Scholes) factoring in their own input of implied volatility and firm-specific adjustments (e.g., early exercise premium, dividend adjustment). This can be done for entire strips across tickers extremely quickly as this is generally run on a Field-Programmable Gate Array (FPGA) — basically, an ultra-high-speed piece of hardware.
However, these thousands of quotes aren’t just set-it-and-forget-it. If you have 1,000 quotes spread across all constituents of the S&P 500, your quotes need to represent the real-time underlying conditions as-is or else you’d just be getting yourself arbitraged until you’re out of the game (e.g., in 5 seconds stock A rockets to $100 and the 99c is still trading for $0.50; a risk free $50 per contract).
Because of this, FIX messages are used to cancel and replace the quotes just as soon as they’re sent out. Take a look:
Sample FIX order + Definitions
8=FIX.4.4|9=237|35=G|49=MarketMaker|56=Exchange|34=987654321|52=20230115-10:15:25.123|11=123456789|41=NewOrderID|55=AAPL|54=1|38=10|40=2|44=2.05|15=USD|21=1|59=0|10=203|
8:FIX version|9:Message length|35:Message type (G for Cancel/Replace)|49:Sender's CompID|56:Target's CompID|34:Sequence number|52:Timestamp|11:Original Order ID|41:New Order ID|55:Symbol (AAPL in this case)|54:Side (1 for Buy)|38:Order quantity (10 contracts)|40:OrdType (2 for Limit)|44:Price (New bid price of $2.05)|15:Currency (USD)|21:TimeInForce (1 for Day)|59:ExpireTime (0 for Good Till Cancel)|10:Checksum
While this may seem like jargon, it represents a simple action: cancel the quote on AAPL at 2.00x2.10, and replace it with a quote at 2.05x2.15. This extremely efficient messaging format makes it possible for quotes to be almost perfectly priced before we even have the time to blink.
That’s the big picture understanding you need, but of course, it always gets deeper:
So, now that we know how some of the major players are making major scratch, let’s look at what they’re trying to go into the new year with. Here’s an excerpt from the job description:
Interestingly, as opposed to their traditional market making practice, this seems to be a pure alpha-generation desk. Just a few weeks prior to this post, we touched upon extracting alpha from gasoline markets with machine learning and positioning data, so this further adds to the idea that there is still alpha to be found in commodity markets without having special advantages (e.g., client flow).
It’s difficult to decipher exactly what the strategy is, but they’ve left yet another clue:
On the same day of the systematic commodities trader posting, a posting for a commodities roll trader was listed. For those unfamiliar, a commodity roll is just when you extend your futures position to the next month. For example, if you own a corn futures contract that expires in December 2023, you would sell it before expiration and buy the contract that expires in March 2023.
Given those 2 pieces of information, it is likely they are developing an alpha that uses fundamental commodity knowledge to predict when commercial traders (e.g., farmers, physical traders), will be rolling out their current positions. Commodity markets are relatively tight, so knowing when the major players are planning to roll creates an opportunity for “front-running” where they can short the front-month in advanced, and simultaneously buy the back-month. Price/volume data can help decipher this as there may be seasonal trends (e.g., at the end of the year from 2017-2022, volume was elevated in the final 2 weeks before December expiration).
While this is speculation, it shows that there are still firms with positive views of creating alpha factories using the traditional measures of being witty with fundamental knowledge and price data. While this may not be their exact strategy, or even close to it, this is an approach we’ll be looking into very shortly.
The Generative AI Problem
It would be a disservice to mention the events of 2023 without mentioning generative AI. So far, the focus on AI has been in the form of large language models (LLMs) like ChatGPT. In the greater finance and banking space, the useful applications are pretty clear-cut and linear; summarize the notes from this recorded client meeting, chatbots on internal company data, pre-fill this compliance document, etc.
However, in the quant space, the power of these systems come from the creative, ultra fine-tuning we’ve come to expect. To see an example of this, let’s look at BondGPT:
As the name suggests, it is essentially ChatGPT but ultra-tuned on corporate bond data. Interestingly, the data is also connected to their liquidity platform that has pricing and depth information on the individual bonds. This might seem pretty vanilla at first, but here at The Quant’s Playbook, we are an enterprising bunch — let’s see how far we can theoretically take this.
Much like with the commodity strategy addressed earlier, we can attack this with a traditional alpha approach. Let’s say we want to run a relative value arbitrage strategy — but first, a quick primer:
Relative Value Arbitrage:
A long/short trade between two bonds with similar characteristics.
For example, assume an AAPL bond is issued at a coupon of 4% and a MSFT bond is issued at a coupon of 3.75%.
Because of market conditions (e.g., over-selling due to fund liquidation), the AAPL bonds trade at a yield of 5%, but the MSFT bonds trade at a yield of 3.5%.
An arbitrageur would buy the AAPL bond, while simultaneously shorting the MSFT bond.
The underlying theory is that over time, the difference in yields between the two bonds will converge back to near its original state.
Ideally, the yield of the AAPL bond will fall back to near 4%, resulting in a higher price (inverse relationship) and profit since the arbitrageur is long the bond
Simultaneously, the yield on the MSFT bonds will ideally increase back to near its original, resulting in a lower price and profit since the arbitrageur is short the bond
Now, understanding the strategy is one part, but just finding suitable opportunities for the strategy has historically been a painful slog. I mean, you first have to filter the bonds by similarity (e.g., industry, credit rating), then you have to deduce why there’s been a divergence, then you have to determine how much liquidity the bonds have, if any.
But now, you can essentially say: “Pull a list of bonds suitable for a relative value strategy with the following criteria: same industry, same credit rating, same maturity year, have at least $10,000,000 in liquidity, and are trading at yields different than their original coupon rate.” You’d not only get suitable pairs in likely a matter of seconds, but also pairs you wouldn’t have considered or discovered on your own.
However, this class of AI is very new and appears to be at the stage where functionality like this is highly experimental. It is entirely possible that these models remain in the niche chat-bot stage and fade into obscurity in the years to come — it’s also entirely possible that these models continue to get stronger and revolutionize the entire quant space (some desk, somewhere must be using them intelligently, right?).
Does Alpha Still Exist?
Yes, yes, and yes.
Quantitative hedge funds are still seeing inflows, pure alpha prop shops are still hiring, and new technology is continuing to enhance old strategies, while crafting new ones.
But we don’t even need to look at that external, anecdotal evidence to support that. Just this year alone, we have cranked out alpha after alpha, ranging from niche event derivatives on alternative markets, to the largest SPX option venues on the planet.
Here’s a look at the most recent performance of one of our live, filtered systems on an alternative market:
To be fair, we’re using extremely powerful models, high-quality data sources, and domain expertise — but still, our resources are dwarfed in comparison to the giant alpha-makers and we’re still able to eek out profits.
Evidently, our small size can be turned into a distinct advantage. Currently, our belief is that the largest opportunity for alpha exists in the smaller, alternative markets that the larger players likely won’t ever touch. Markets like Kalshi and Polymarket that offer quirky bets on the S&P 500, fed rate cuts, and just general things in the news like politics and pop culture.
Extracting alpha from these niche markets are particularly attractive for a few reasons:
We end up with proprietary alternative datasets.
For instance, if we wanted to tackle the popular market of the high temperature in NYC (e.g., 47 to 48), we would use our skillset to build a training dataset of aggregate forecasts with typical deviations, and a few other features to create a robust model for predictions.
This gives us a further edge over the general betting amateurs as we are able to have a better quantitative gauge of probabilities.
They actually offer pretty good returns.
Referencing our S&P 500 model set and strategy, each trade offers us on average 20-30% return on a consistence, near-daily basis, which is almost unheard of in traditional equity/derivative markets. The option market equivalent of our strategy routinely offers 5%, albeit with a significantly worse risk profile.
So far we’ve only tackled a few sub-markets on these platforms, but there is upcoming research that will legitimately shock you as to just how much is up for grabs.
All in all, the outlook for quantitative finance has has never looked better. There is still juice in the old strategies, and there’s more than ever in the new ones.
If you’re reading this publication, you’re already a step ahead of the pack. 2023 was our first full year of weekly novel research and we’ve come along way. I am deeply grateful for you, as without your interest, making these strides would be infinitely times more difficult, if not impossible.
As a continued show of gratitude, we will be entering the new year with the same high-quality standards you’ve come to expect, building on the research of yesteryear for truly powerful and proprietary insights and strategies you can only get here.
Here’s to one awesome year, and many more to come. 🍻🍻
Happy trading! 😄
Hi, what's the model generated above pnl chart? is it a tree model or arimax model?
Hey there, I'm having financial troubles currently and I was wondering if you could cancel and refund my subscription, I lost my job earlier in December and I forgot to cancel this subscription and not having the money would make even buying food difficult.
Ill resubscribe once I have a job.
Thank you.